WITHDRAWN: Integrated feature selection (IFS) algorithm and enhanced weight based convolutional neural network (EWCNN) for social emotion classification

Abstract

Abstract To design and implement IFS-EWCNN for improve the performance of social emotion classification for twitter reviews. Design a new Integrated Feature Selection (IFS) algorithm, which combines filter and wrapper methods procedure for evaluating semantic gap among source features produced using teaching model and target task characteristics before the application of sparse encoding to transfer learning. Filtering algorithm follows the procedure of the information gain and Pearson’s Correlation. In the wrapper selection algorithm follows the procedure of the Expectation Maximization with Forward Search (EMFS), and Exhaustive feature selection. Then implemented a novel Animal Migration Optimization based Sparse Encoding (AMO-SC) for forming dense high-level features by transforming sparse low-level features and further investigation is required on effectiveness of this method on emotion classification. Finally, for classifications of emotions, proposed an Enhanced Weight based Convolutional Neural Network (EWCNN). It concludes that proposed IFS-EWCNN classifier achieves better performance compared with the existing methods in terms of f-measure, specificity, sensitivity, recall, precision and accuracy. The experiments are carried out in SemEval 2016 Task 4A for sentiment and SemEval 2018 Task 1C emotion classification using Matlab tool. The proposed IFS-EWCNN classifier approaches accurately classify the social emotions.